Chapter 10: Patient-reported Health Instruments: Carer Impact

Transcription

1 Chapter 10: Patient-reported Health Instruments: Carer Impact The impact of a disease on a patient is an increasingly important outcome measure in medicine and healthcare. Issues such as quality of life are now widely used in clinical trials and in patient management for assessing morbidity and the impact of treatment. For a long time, studies focused almost exclusively on changes in the impact of ill health on patients, but increasing attention is now being paid to the impact on carers of patients with chronic diseases. Carers (or caregivers) play an important role in the care of chronically ill patients, as the number of people with chronic illnesses is increasing and informal and community care outside of acute services is increasingly encouraged. Carers tend to be family members (often the spouse) or friends, who are called upon to provide significant and continuous support to the person with ill health. It is being increasingly recognized that caring for someone with ill health poses challenges and can represent a stressful and difficult situation to the carer with adverse physical and psychological outcomes for the caregiver. Two different approaches have been used to study carer burden. The first approach, which is an indirect approach, uses generic instruments as proxy measures such as the SF-36. Generic instruments have usually been extensively tested, although not necessarily in the carer population. They provide a broad perspective of health, but they do not give an insight into carers specific problems. The second approach, which is direct, investigates the carers experience, focusing specifically on the content of carers experiences. It uses either instruments that have been developed for carers generally (hereafter referred to as general carer instrument) or instruments that have been developed for people caring for a person with a specific condition (hereafter referred to as disease-specific carer instrument). An example of a general carer instrument is the Carer Strain Index. These instruments provide a more specific measure of the carer-specific burden. However, these instruments are not specific to a particular disease group, and as such may not capture all the relevant issues for a person caring for a patient with a particular condition. Thus, a number of diseasespecific carer instruments have also been developed, for example, the Parkinson s Impact Scale (PIS). Instruments that have been used to assess carer burden also vary in terms of their dimensions, with some instruments investigating multiple dimensions (including for instance, physical health, psychological health, social roles), and other instruments being dimension-specific (e.g. fatigue, depression). Caregiver well-being has traditionally been considered from a deficit perspective and little attention has been paid to positive aspects of caregiving (Berg-Weger and Tebb 1998) and increasingly some instruments also focus on positive aspects of caregiving. This review reports the psychometric properties of generic and carer-specific instruments that have been used in people who care for people with ill health. This review does not discuss dimension-specific or disease-specific instruments. Furthermore, the focus of this review is on caregiving for adults with ill health, not for children (either healthy or with ill health) or healthy elderly. The review only includes articles published in English with data from English speaking populations (UK,, Canada, Australia and New Zealand). 349

2 RESULTS: Patient-reported Health Instruments: Carer impact Search terms and results: identification of articles At the time of the review, the PHI database contained 12,000+ records (up to June 2005). Search results are detailed in table When assessed against the review inclusion criteria, 44 articles were retrieved and reviewed in full. Of these, 26 articles were included in the review. Table 10.1 Source Results of search No. of articles considered eligible Number of articles included in review PHI database: original search (up to June 2005) Total number= Supplementary search TOTAL Supplementary searches included hand-searching of titles from 2004 to 2006 of the following key journals: -Health and Quality of Life Outcomes -Medical Care -Quality of Life Research Further searches were conducted within the bibliography and using Pub Med per instrument up to September Identification of instruments Five indirect measures in the form of generic health instruments were included in the review, together with 26 general carer instruments. The developmental and evaluative studies relating to the generic health instruments reviewed are listed in Tables 10.2 to Those relating to general carer instruments are shown in Tables 10.7 to Table includes examples of carer disease-specific instruments. 350

3 RESULTS: GENERIC INSTRUMENTS (INDIRECT APPROACH) Five generic instruments were identified which were evaluated for use to assess carer impact. Full details of the development, domains and scoring methods are detailed in Chapter 3. The following instruments measurement properties are reported: a) SF-36 and SF-12 b) Health Utilities Index Mark 2 c) Reintegration to Normal Living Index d) Ferrans and Power Quality of Life Index e) General Health Questionnaire a) SF-36 and SF-12 Reliability Good internal consistency for the SF-36 overall was reported in studies by Jenkinson et al., (2000); Berg-Weger et al., (2003) and for the Physical Component Subscale (PCS) and Mental Component Subscale (MCS) in a study by Clark et al., (2004). One study found adequate internal consistency for the other subscales (Berg-Weger et al., 2003)., whereas another found good internal consistency for the different SF-36 subscales (Cameron et al., 2006b). The SF-12v2 has been found to have weak internal consistency for a sample of carers of dementia patients (McConaghy and Caltabiano 2005). Validity Convergent and discriminant validity Depression, measured by the Centre for Epidemiologic Studies-Depressed Mood Scale, was found to be significantly related to the Physical Health and Mental Health domains of the SF-36 (Berg-Weger et al., 2003). In the same study, anxiety (assessed by the Self-rating Anxiety Scale) was also significantly negatively related to Physical Health and Mental Health, and physician s visits. Visits to mental health professionals were only significantly and negatively related to The Mental Health summary score. There was also moderate discriminative validity, as alternative mental heath measures correlated more strongly with the Mental Health subscale than with the Physical Health subscale. On the other hand, alternative measures of physical health correlated more strongly with the Physical Health subscale. Internal validity Factorial analysis supported the original structure of the SF-36 in a study by Berg- Weger et al., (2003). Predictive validity One study found support for the predictive validity of the SF-36, with particularly the Vitality Scale being a predictor of stroke carer stress (Smith et al., 2004). 351

4 Socio-demographic variables The scores on the SF-36 for carers have been found to be below those for the general population in studies by Jenkinson et al., (2000) and Cameron et al., (2006b). The scores on the SF-12 have also been found to be slightly below the general norms (Clark et al., 2004). In a further study, carer scores were reported to be lower than population norms on the Energy and Vitality scales (Smith et al., 2004). Generic carer instruments The PCS and MCS has strong correlations with the Caregiver Strain Index (CSI), supporting construct validity in a study by Jenkinson et al., (2000). Responsiveness No data available. Precision Some floor effects have been found for the SF-36 for carers for the Role Physical (19.3%) and Role Mental (23.4%), as well as some ceiling effects (50.6% and 49.0% respectively) (Jenkinson et al., 2000). Acceptability No data available. Feasibility The survey was administered by telephone interviews in a study by Berg-Weger et al., (2003), reporting 30 minutes completion time. 352

5 Table 10.2: Evaluative studies relating to the SF-36 when completed by carers of people with ill health Study/ Country Population (N) Age (years) Method of administration Setting Measurement and Practical properties SF-36 Reliability Validity Responsiveness Precision Acceptability Feasibility Berg-Weger et al., 2003 Clark et al., 2004 Smith et al., 2004 UK Jenkinson et al., 2000 Multi-national including the UK Cameron et al., 2006b Adult daughters who had been primary caregivers to a parent with Alzheimer Disease or a related disorder, who had died at least 12 months before the study (102) Age 57 years Telephone interview Alzheimer Association members Family caregivers of stroke survivors (132) SF-36v2 mental and psychical scales Age 56.7 Sampled from a sample of a national, multi-site clinical trial Interviewer administered Carers of stroke patients (90) and stroke patients Age 57.8 Patients identified from 2 hospital stroke registers Carers of patients with amyotrophic lateral sclerosis (415) Age 55.1 years Carers of patients recruited through 74 clinical sites throughout Europe Self-administered Caregivers of patients with acute respiratory distress syndrome Factorial validity Convergent and discriminant validity Predictive Canada SF-12v2 McConaghy and Caltabiano 2005 Carers of people with dementia (42) Age 62.0 years Self-completion questionnaire Self-administered or face to face interviews Australia 353

6 b) HUI 2 Reliability No data available. Validity One study evaluated criterion validity of HUI 2 in carers of patients with Alzheimer Disease (Bell et al., 2001) by comparing HUI 2 to a caregiver time questionnaire, a caregiver burden instrument and the SF-36. It was found that the HUI 2 may not adequately capture differences in the burden of caregivers for patients with Alzheimer Disease. Responsiveness/ Precision/ Acceptability/Feasibility No data available. Table 10.3: Evaluative studies relating to the HUI 2 when completed by carers Study/ Country Population (N) Age (years) Method of administration Setting Measurement and Practical properties Bell et al., 2001 Canada 679 caregivers of individuals with Alzheimer Disease 63 years Interviewer administered questionnaire Reliability Validity Responsiveness Precision Acceptability Feasibility c) RNLI Reliability Good internal consistency has been found for the RNLI in a study by Bluvol and Ford-Gilboe (2004). Validity/ Responsiveness / Precision/ Acceptability No data available. Feasibility The questionnaires (the RNLI, which has 11 items, plus 2 more questionnaires) took minutes to complete in a study by Bluvol and Ford-Gilboe (2004). 354

7 Table 10.4: Evaluative studies relating to the RLNI when completed by carers Study/ Country Population (N) Age (years) Method of administration Setting Measurement and Practical properties Bluvol and Ford- Gilboe 2004 Carers (40) and stroke patients (40) Age carers 66.2 Self-completion questionnaire Reliability Validity Responsiveness Precision Acceptability Feasibility Canada d) FPQLI Reliability High internal consistency was found for the Total instrument and moderate to high alphas for the Life domains (Weitzner et al., 1997). Validity Socio-demographic variables Caregiver age was significantly correlated with the Health/functioning and Psychological/spiritual domains, as well as the Total score in a study by Weitzner et al., (1997). Responsiveness/ Precision/ Acceptability/ Feasibility No data available. Table 10.5: Evaluative studies relating to the FPQLI when completed by carers Study/ Country Population (N) Age (years) Method of administration Setting Measurement and Practical properties Weitzner et al., 1997 Caregiver of cancer patient (22) Age51.7 Interviewer administered Reliability Validity Responsiveness Precision Acceptability Feasibility 355

8 e) GHQ Reliability No data available Validity Discriminative validity There was a non-significant trend in GHQ Total scores and Depression subscales scores to be higher for carers using Admiral Nurse (AN) teams vs. carers who did not (Woods et al., 2003). On follow-up, a significant difference was found on the Anxiety and Insomnia subscale, where outcome was better for the AN group. Another study showed that carers of dementia patients showed higher levels of distress as measured by GHQ than carers for patients with depression (Rosenvinge et al., 1998). Furthermore, significant differences in GHQ scores have been found between carers of people with anorexia and psychosis (Treasure et al., 2001). GHQ scores have also been found to differ in carers of people with a head injury according to different time intervals post-injury. The GHQ scores were higher for carers of people with a recent head injury, which indicates greater burden in this group (Sander et al., 1997). Predictive validity Coping style has been found to contribute significantly to GHQ score variance, with emotion-focused coping being related to GHQ scores in a study by Sander et al., (1997). Furthermore, coping accounted for more of the GHQ variance than disability scores. Socio-demographic variables Gender has been found to have a significant effect on GHQ scores, but neither race nor relationship to the injured person had a significant effect (Sander et al., 1997). Dimension-specific variables Strong positive correlations were found between the GHQ and the Relatives Stress Scale (Draper et al., 1992). Responsiveness The GHQ-28 has been shown to be responsive to change in a study using cognitive behavioural therapy in carers of Parkinson s disease patients. Both the Total score and the scores for 3 of the sub-scales decreased in response to the intervention (Secker and Brown 2005). Both conventional and AN services led to lower GHQ scores overall and 2 of the 4 subscales over an 8-month period (Woods et al., 2003). Precision/ Acceptability/ Feasibility No data available. 356

9 Table 10.6: Evaluative studies relating to the GHQ when completed by carers Study/ Country Draper et al., 1992 Australia Sander et al., 1997 Rosenvinge et al., 1998 UK Treasure et al., 2001 UK Woods et al., 2003 UK Secker and Brown 2005 UK Population (N) Age (years) Method of administration Setting Co-resident carers of dementia (51) and stroke patients (48) Age stroke 72.0 and dementia 76.0 Self-completion Carers of patients with head injury (60) 3 groups corresponding to 3 postinjury intervals: early (26), intermediate (21) and long-term (22) Age early 39.1, intermediate 45.8 and late 45.1 Self-completion Carers of patients with dementia (32) or depression (25) Age 67.8 for dementia carers 66.8 for depression carers Interviewer administered Carers of patients with anorexia (71) or psychosis (68) Interviewer administered Carers of people with dementia (128, of which 55 used an admiral nurse (AN) service and 73 did not (comparison group)) Age 62.4 for AN group and 58.8 for comparison group Interviewer administered Carers of patients with Parkinson s (30) Age 59.1 for treatment group (n=15) and 58.8 for control group (n=15) Measurement and Practical properties Reliability Validity Responsiveness Precision Acceptability Feasibility 357

10 RESULTS: DIRECT MEASURES OF CARER IMPACT Eight instruments that investigate the carers general burden have been identified. Full details of the development, domains and scoring methods are detailed in Tables 10.7 and The following instruments measurement properties are reported: a) Appraisal of Caregiving Scale b) Bakas Caregiver Outcomes Scale c) Caregiver Burden Inventory d) Caregiveing Burden Scale e) Caregiver Impact Scale f) Caregiver Strain Index g) Caregiver Well-Being Scale h) Zarit Burden Interview a) Appraisal of Caregiving Scale (ACS) The ACS has been developed in the with carers of cancer patients receiving radiotherapy. The ACS is a 53-item instrument that measures the meaning of illnesscaregiving situation in terms of the intensity of four dimensions (Harm/loss, Threat, Challenge and Benign). b) Bakas Caregiver Outcomes Scale (BCOS) The BCOS was developed to measure changes in caregiving outcomes. The BCOS is a unidimensional scale based on 10 items and addresses changes in caregiving social functioning, subjective well-being and physical health. It was first developed and evaluated in carers of stroke survivors in the. c) Cargiver Burden Inventory (CBI) The CBI is a 25 item instrument with 5 subscales that was developed in carers of confused or disoriented older people in Canada. The CBI aims to give a reading of caregivers feelings and a picture of the carers responses to the demands of caregiving. d) Caregiver Appraisal Scale (CAS) The CAS was designed as a 47-item interview questionnaire for caregivers of disabled elderly. The CAS has five domains of caregiving appraisal: Caregiving satisfaction, Perceived caregiving impact, Caregiving mastery, Caregiving ideology and Subjective caregiving burden. 358

11 e) Caregiver Impact Scale (CIS) The CIS was developed in Canada in carers of cancer patients and is based on the Illness Intrusiveness Rating Scale by (Devins et al., 1983). The CIS assesses the extent to which providing care interfered with the caregiver s participation in 14 domains of lifestyle (such as health, employment, recreation). f) Caregiver Strain Index (CSI) The CSI was developed in the to measure caregiver reactions, including perception and emotional feeling with regards to caregiving. It includes 13 items and was developed in caregivers of patients, aged 65 or over, who had recently been hospitalized for hip surgery or heart problems. At a later stage it was modified, to include a sometimes response category, rather than just the yes/no response options and some items were rephrased. g) Caregiver Well-Being Scale The Caregiver Well-Being Scale was developed in the in a sample of caregivers of older adults with dementia, caregivers of children with developmental problems and caregivers of healthy children who were younger than 12 years of age. The scale has also been evaluated in a sample of caregivers of chronically ill patients. It includes 45 items, with 2 subscales: Basic Human Needs and Activities of Daily Living. The scale identifies the caregivers strengths in meeting their basic needs and daily activities. h) Zarit Burden Interview (ZBI) The Zarit Burden Interview assess the degree to which a caregiver perceives their caregiving responsibilities to have a negative effect on their health, personal and social life, finances and emotional well-being. Different versions of the ZBI are available, the 22-item full version and the 12-item short version. The ZBI was developed and tested in carers of patients with dementia in the. 359

12 Table 10.7: General carer instruments Instrument (no. items) Domains (no. items) Response options Score Administration/ Completion (time) Appraisal of Caregiving 5 point Likert Scale Self-completion Scale (53) 1= very untrue, 5= very true 20 minutes Bakas Caregiving Outcomes Scale (10) Caregiver Burden Inventory (24) 4 domains harm/loss (15 items) threat (15 items) challenge (15 items) benign (8 items) Unidimensional 5 domains: developmental, physical, social, emotional burden and time dependence 7-point Likert scale 1=changed for the worst, 7=changed for the best 5 point Likert scale 0=strongly disagree, 4=strongly agree Caregiving Impact Scale 14 domains of caregivers lifestyle 7-point Likert scale 1=not very much, 6=a lot Caregiving Appraisal Scale (original 47, later 35) Carer Strain Index (CSI) (13 items) Caregiver Well-Being Scale (45 items) 5 domains (47 items): caregiving satisfaction, perceived caregiving impact, caregiving mastery, caregiving ideology and subjective caregiving burden. 4 domains (35 items): perceived burden (15), caregiver relationship (11) satisfaction, caregiver ideology (5) and caregiving mastery (4) Self-completion questionnaire uses 5 point scale 1= strongly disagree and 5=strongly agree Higher scores on each subscale represent greater intensity of the appraisal dimension Items for each domain are summed. Domain scores range from 0-20, except for physical burden (0-16). For physical burden the summed score is multiplied by 1.25 to give an equivalent score out of 25 Summation of items with higher scores indicating higher interference For 35 item instrument: Scores calculated by summing individual domain scores, using reversed scoring for certain items. Unidemensional (lowest to highest level of strain) Basic human needs (4 factors and 22 items) Activities of daily living (5 factors and 23 items) Self-completion or interviewer administered Self-completion or interviewer administered Self-completion or interviewer administered Interview or self-completion Self-completion or interviewer administered minutes Self-completion Zarit Burden Interview (Original 29, full version 22 items, short version 12 items) For 12 item short version Personal strain (9) Role strain (3) 5-point Likert style 0=never and 4=nearly always 0-88 with higher scores a greater burden Interviewer administered 360

13 Table 10.8 summarizes the domains included in the different instruments. However, some general carer instruments do not include these domains as such, but include items that reflect these domains. For example, the CSI is unidimensional. Another example is the CAS which does not include a social well-being domain, but within the perceived burden domain several items relate to social well-being such as my social life has suffered or I feel isolated and alone. Other domains found in general caregiving instruments are not reflected in the health status domains by Fitzpatrick et al. (1998), such as the domain of caregiving mastery of the CAS, which includes items on how well the carer copes with caring or time-dependence burden of the Caregiving Impact Scale, which describes burden due to restrictions of caregivers time. Also, some domains (symptoms, cognitive function and treatment satisfaction) from Fitzpatrick et al. (1998) are of less relevance for carer instruments. Table 10.8: Summary of carer-specific instruments: health status domains (after Fitzpatrick et al., 1998) Instrument Appraisal of Caregiving Scale Bakas Caregiving Outcomes Scale (10) Caregiving Appraisal Scale (47/35) Caregiver Burden Inventory (24) Caregiving Impact Scale (14) Carer Strain Index (13) Caregiver wellbeing scale (45) Zarit Burden Interview (29, 22, 12 or 4) Physical function Symptoms Global judgement Psychol. wellbeing Instrument domains Social wellbeing Cognitive functioning Role activities x x x x x x x x x x Personal construct x x x x x x x x x x x x x x x x x x x Treatment satisfaction 361

14 RESULTS: CARER IMPACT a) Appraisal of Caregiving Scale (ACS) Reliability Two studies have found the ACS sub-scales to be internally consistent (Oberst et al., 1989; Carey et al., 1991). Validity Strong correlations were found between the Harm/loss and Threat subscales, indicating that they may represent the same construct (Oberst et al., 1989). Also, the high correlation between the challenge and benign subscales represent a problem. Socio-demographic variables Each of the four sub-scales was related to at least one other caregiver variable (Oberst et al., 1989). Harm/loss scores were correlated with the carer s level of education, social status and health status. Threat scores were correlated with the carer s level of education and social status. Challenge and benign scores were correlated with caregiver age. Benign scores were related to the carer s perception of the illness (as more or less serious). Patient variables Correlations were also found with various patient variables (Oberst et al., 1989). Harm/loss scores for carers were related to the length of time patients received radiation. The carer s relationship to the patient was also related to the benign subscale, with those caring for a parent perceiving the situation as less benign than those caring for a spouse or others. Table 10.9: Developmental and evaluation studies relating to the Appraisal of Caregiving Scale: Study/ Country Oberst et al., 1989 Carey et al., 1991 Population (N) Age Method of administration Setting Family members of cancer patients receiving radiotherapy.(47) Age 53.3 Self-completion Family caregivers of patients receiving outpatient chemotherapy (49) Self-completion Measurement properties Reliability Validity Responsiveness Precision Acceptability Feasibility 362

15 b) Bakas Caregiver Outcomes Scale Reliability Good internal consistency was found for both the 10 and 12 item BCOS in a study by Bakas and Champion (1999). High internal consistency for the 10-item BCOS was also found in two further studies (Bakas et al., 2004; Bakas et al., 2006). One study reported finding good test-retest reliability (0.66 for the 15-item BCOS and 0.68 for the 10-item BCOS) (Bakas et al., 2006). It was not clear whether the findings of testretest reliability refer to group or individual comparisons. Validity Internal validity Two studies report evidence on construct validity, by using factor analysis that supported unidimensionality of the BCOS (Bakas and Champion 1999; Bakas et al., 2006). Generic health status Significant weak to moderate correlations with LIFE-3 and with the SF-36 subscales were found in a study by Bakas and Champion (1999) and a significant weak correlation was found with the SF-36 General Health Subscale by Bakas et al., (2006). Responsiveness/ Precision/ Acceptability/ Feasibility No data available. Table 10.10: Developmental and evaluation studies relating to the Bakas Caregiving Outcomes Scale: Study/ Country Bakas and Champion, 1999 Bakas et al., 2004 Bakas et al., 2006 Population (N) Age Method of administration Setting Caregivers of stroke survivors (sample 1= 92, sample 2= 104) Age sample and sample Self-completion questionnaire Caregivers of stroke survivors (114) Age 60.5 Self-completion questionnaire Family caregivers of stroke survivors (147) () Age 51.7 years Interviewer administered (face to face or telephone) or self-completed Measurement properties Reliability Validity Responsiveness Precision Acceptability Feasibility Internal consistency Test-retest Construct 363

16 c) Caregiver Burden Inventory (CBI) Reliability The total CBI score (Foster and Chaboyer 2003) and the five subscales of the CBI have been found to be internally consistent in studies by Novak et al., (2001); and Foster and Chaboyer (2003). Validity Internal validity The five factor structure was supported empirically in a study by Novak and Guest (1989). Dimension-specific variables The total CBI score, as well as 4 of the 5 subscales (with the exception of emotional burden), have been found to be significantly correlated to filial (family) obligation (Foster and Chaboyer 2003). Responsiveness The Total CBI score was responsive to change in carers of patients with Parkinson s disease receiving cognitive behavioural therapy after 3 months of therapy, compared to controls in a study by Secker and Brown (2005). Precision/ Acceptability/ Feasibility No data available. 364

17 Table 10.11: Developmental and evaluation studies relating to the Caregiving Burden Inventory: Study/ Country Novak and Guest, 1989 Canada Foster and Chaboyer, 2003 Australia Secker and Brown, 2005 UK Population (N) Age Method of administration Setting Carers of confused or disoriented older people (107) Age 60.1 Interviewer administered Carers of a family member who had been critically ill & admitted to intensive care (71) Age 50.3 Self-completion Carers of patients with Parkinson s (30) Age 59.1 for treatment group (n=15) and 58.8 for control group (n=15) Measurement properties Reliability Validity Responsiveness Precision Acceptability Feasibility d) Caregiving Appraisal Scale Reliability Internal consistency was good for the three factors of the ACS for 2 different samples of carers (Lawton et al., 1989). However, it has to be noted that the 3 factors were different for the 2 samples. Internal consistency was good for three of the four factors (Struchen et al., 2002). Caregiving mastery showed poor internal consistency, but since the factor emerged from the analysis of 2 different cohorts, Struchen et al. (2002) believed that it was a significant construct of caregiver appraisal. Using two of the original subscales (perceived burden and impact of caregiving role), Dracup et al. (2004) found good internal consistency of these subcscales. Test-retest reliability was found to be reasonable in one of the samples investigated by Lawton et al (1989). Validity Construct Validity Factor analysis has provided conflicting evidence of the structure of the CAS. Three factors found in a sample of carers of disabled elderly people in respite were not confirmed in a sample of carers of disabled elderly people in institutions (Lawton et al., 1989). Another study found a five-factor solution, but two factors (caregiving mastery and caregiving burden) were found to be less robust (Lawton et al., 1991). Another study concluded that factor analysis showed that the CAS has four subscales: perceived burden and caregiver satisfaction with their relationship to the patient, caregivers ideology and caregiving mastery (Struchen et al., 2002). The four 365

18 factor solution was also found with 35 items (rather than the initial 47) and in different samples (carers of the traumatic brain injury model system and carers of the residential treatment programme cohort) (Struchen et al., 2002). Generic health status One of the factors (perceived burden) was significantly correlated to the GHQ, Subjective Burden Scale and Objective Burden Scale (Struchen et al., 2002). A negative significant correlation was found between caregiver relationship satisfaction and the Objective Burden Scale. However, this was a weak correlation. Dimension-specific variables The caregiving satisfaction subscale of the ACS was found to be significantly related to caregiving burden and caregiving burden was related to depression (Lawton et al., 1991). Validity/ Responsiveness/ Precision/ / Feasibility No data available. Acceptability Of 241 participants, 11 cases had more than one response missing (Struchen et al., 2002). Table 10.12: Developmental and evaluation studies relating to the Caregiving Appraisal Scale: Study/ Country Lawton et al., 1989 Lawton et al., 1991 Struchen et al., 2002 Dracup et al., 2004 Population (N) Age Method of administration Setting Carers of disabled older people in respite care or in institutions (632) Age 59.7 respite care, 76.2 institutionalized care Interviewer administered Spouse (285) and adult child (244) carers of elderly people with Alzheimer s Disease Age 76.3 Interviewer administered Carers of person with traumatic brain injury (241) Age47.0 Self-completion Spouses of patients with heart failure (75) Age 54.0 Self-completion Measurement properties Reliability Validity Responsiveness Precision Acceptability Feasibility 366

19 e) Caregiving Impact Scale Reliability The CIS has been found to be internally consistent in two studies (Cameron et al., 2002; Cameron et al., 2006a). Validity/ Responsiveness/ Precision/ Acceptability/ Feasibility No data available. Table 10.13: Developmental and evaluation studies relating to the Caregiving Impact Scale: Study/ Country Population (N) Age Method of administration Setting Measurement properties Caregiving Impact Scale Reliability Validity Responsiveness Precision Acceptability Feasibility Cameron et al., 2002 Canada Family caregivers of cancer patients (44) Age 55.8 Interviewer administered Cameron et al., 2006a Canada Informal carers of stroke survivors (94) Age 60.8 Interviewer administered or self-completion f) Caregiver Strain Index Reliability Internal consistency Four studies found the CSI Total score to be internally consistent (Robinson 1983; Berg-Weger et al., 2000; Jenkinson et al., 2000; Diwan et al., 2004). However, one study found internal consistency for 2 of the 3 subscales below 0.7 (Diwan et al., 2004). The modified CSI has also been found to be internally consistent (Thornton and Travis 2003). Test-retest reliability Test-retest reliability had not been investigated in the original CSI studies. However, in the modified CSI study, it was found that test-retest reliability was better for the modified CSI than the parent CSI (Thornton and Travis 2003). 367

20 Validity Construct validity Principle component analysis identified 3 factors of the CSI, which were comparable but not identical to findings reported by others (Diwan et al., 2004). Exploratory factor analysis and structural equation modelling found three factors that the authors reported to be similar, but not identical, to those proposed originally by Robinson in 1984 (Rubio et al., 1999). However, it is unclear what the authors are referring to, as Robinson (1984) does not report any factors or sub-scales of the CSI in the original development. Socio-demographic variables No significant difference was found in the level of strain as measured by the CSI at three or six months after a stroke between men and women in a study by Blake et al., (2003). Adult children were significantly more likely to report role strain compared to spouses and other carers. Higher income of the caregiver was predictive of greater role strain, and perceived lack of support from health care services was associated with greater personal strain (Diwan et al., 2004). For the modified CSI, age was found to be inversely related to carer strain (Thornton and Travis 2003). Generic health status The CSI has been found to be significantly moderately to strongly correlated with the General Health Questionnaire-12 (Blake and Lincoln 2000; Blake et al., 2003), patient Extended Activities of Daily Living Scale (EADL), and Negative Affectivity (Blake et al., 2003). In one study, the best predictor of the CSI was the carer s mood and other factors were the perceived patient EADL and negative affectivity (Blake and Lincoln 2000). In a second study, strain was accurately predicted by a model based on the General Health Questionnaire-12, Positive and Negative Affectivity Schedule (Blake et al., 2003). Also, CSI scores correlated moderately with PCS and weakly with MCS scores of the SF-36 in a study by Jenkinson et al., (2000). Patient variables Significant correlations were found between the CSI and a variety of patient variables (Robinson 1983). Positive correlations were found for CSI score and the patient s age, re-hospitalization within to months and mental status. Negative correlations were found with the patients ability to perform activities of daily living and satisfaction with progress during convalescence. Another study however, found no significant correlation was found for CSI score with age of the patient or time since the stroke of the patient (Blake and Lincoln 2000). The modified CSI was found to be significantly correlated to the patient s mental capacity and physical functioning and the patient s age (Thornton and Travis 2003). 368

21 Caregiver variables CSI scores were correlated with a number of variables of caregivers perceptions (for example carer s perception that they were very involved with caregiving or emotional strain of the caregiver) (Robinson 1983). Responsiveness CSI scores have been shown to significantly reduce in a study examining effectiveness of cognitive behavioural therapy for the carer (Secker and Brown 2005). Feasibility Interviews took minutes in a study by Diwan et al., (2004), but this included the completion of several other questionnaires. For the modified CSI, interview time was between 10 and 20 minutes (Thornton and Travis 2003). Precision/ Acceptability No data available. 369

22 Table 10.14: Developmental and evaluation studies relating to the Carer Strain Index: Study/ Country Population (N) Age Method of administration Setting Measurement properties Carer Strain Index Reliability Validity Responsiveness Precision Acceptability Feasibility Robinson, 1983 Carers of recently hospitalized hip surgery or heart patients (81) Age Interviewer administered Rubio et al., 1999 Caregivers of adults with dementia (27) and children with developmental problems (8) or healthy children younger than 12 years of age (53); and non-caregivers (77) Self-report questionnaire Blake and Lincoln, 2000 UK Co-resident spouses of stroke patients (222) Age 69.0 Self completion questionnaire Berg-Weger et al., 2000 Jenkinson et al., 2000 Europe incl. UK Blake et al., 2003 UK Thornton and Travis (2003) Modified CSI Diwan et al., 2004 Secker and Brown, 2005 UK Caregivers of chronically ill family members (142) Self-report Recruited through six caregiver-support organizations Carers of patients with amyotrophic lateral sclerosis (415) Age 55.1 years Set in 74 clinical Self-completion questionnaire Spouses of stroke patients (130 at 3 months and 116 at 6 months) Age 66.4 Caregivers of family members or friends aged 53+ taking medication regularly and receiving formal or informal assistance (158) Interviewer administered Caregivers of community-dwelling dementia patients (150) Age 61.9 Interviewer-administered Carers of patients with Parkinson s (30) Age 59.1 for treatment group (n=15) and 58.8 for control group (n=15) 370

23 g) Caregiver Well-Being Scale Reliability Internal consistency was high for the scale overall, for the two subscales, and for the four factors of the first subscale (Basic Human Need) and for three out of the five factors of the second subscale (Activities of Daily Living) (2 factors had low internal consistency) (Tebb 1995). The Caregiver Well-Being Scale had good reliability (Berg-Weger et al., 2000). Validity Face validity Face validity was examined by four people familiar with the caregiving literature (Tebb 1995) Construct validity For construct validity, moderate to high correlations were found to the Computerized Stress Inventory (Tebb 1995). Lifestyle satisfaction scores were moderately to highly correlated to the two sub-scales of the Caregiver Well-Being Scale (Tebb 1995). Factor analysis suggested that some items could be deleted from the questionnaire, as these items did not load highly on any factor (Berg-Weger et al., 2000). Furthermore, structural equation modelling showed that the original models did not fit the data, and consequently the models were revised to fit the data (Rubio et al., 1999). For example, only three of the original four constructs measured the sub-scale of basic needs. Criterion validity Criterion validity was assessed by comparing the scores of caregivers to noncaregivers. Not all the expected significant differences were found, but the differences were in the expected direction (Tebb 1995). Responsiveness/Precision/Acceptability/Feasibility No data available. 371

24 Table 10.15: Developmental and evaluation studies relating to the Caregiver Well- Being Scale: Study/ Country Population (N) Age Method of administration Setting Measurement properties Caregiver Well-Being Scale Reliability Validity Responsiveness Precision Acceptability Feasibility Tebb, 1995 Rubio et al., 1999 Berg- Weger et al.,2000 Caregivers of adults with dementia (27) and children with developmental problems (8) or healthy children younger than 12 years of age (53); and noncaregivers (77) Self-report questionnaire Re-analysis of the sample from the study by Tebb (1995) Caregivers of chronically ill family members (142) Self-report Recruited through six caregiversupport organizations Internal consistency h) Zarit Burden Interview Reliability Face Criterion and construct Good internal consistency was found for the 29-item ZBI (Williams 1993), the 22- item ZBI (Zarit et al., 1987; Whitlatch et al., 1991; Majerovitz 1995; Hebert et al., 2000; Bedard et al., 2001; McConaghy and Caltabiano 2005), for a short (12-item) ZBI (Hebert et al., 2000; Bedard et al., 2001; O'Rourke and Tuokko 2003) and for a 4- item screening ZBI (Bedard et al., 2001). Good internal consistency has also been found for the 2 sub-scales (Personal Strain and Role Strain), as well as individual items of the ZBI (Whitlatch et al., 1991). Validity Internal validity Factor analysis supported a two factor solution for the short ZBI (Hebert et al., 2000; Bedard et al., 2001; O'Rourke and Tuokko 2003), but the short ZBI showed a better adjustment than the original ZBI (Hebert et al., 2000). Strong significant correlations were found between the original ZBI and the short ZBI (Bedard et al., 2001), making the short version of the ZBI comparable to the full version. 372

25 Socio-demographic variables For the 29-item ZBI, no difference was found in the total score between daughters and spouses as caregivers (Zarit et al., 1980). However, in a later study, daughters and wives were found to score significantly higher on the ZBI than other caregivers (Zarit et al., 1987). Significant differences in ZBI score (20-item) were found between husband and wife carers (Zarit et al., 1986). No significant effect was found for education and income (Zarit et al., 1987). Scores for the short ZBI (12-item) were significantly higher for women than for men (Bedard et al., 2001). However, another study did not find any correlation between ZBI score and the gender of the caregiver (Hebert et al., 2000). Furthermore, the same study did not find any correlation of the ZBI score and marital status and employment status. It has also been found that more women than men have a higher ZBI score (Gallicchio et al., 2002). Younger carer age has also been found to be significantly associated with a higher ZBI score (Schneider et al., 1999) Patient variables Contrary to expectations, none of the variable on the patient s impaired behaviours were correlated with the carer s burden (Zarit et al., 1980). Significant associations were found between ZBI scores and behavioural disturbance (behavioural deficits) and cognitive impairment of the patient (Schneider et al., 1999). No correlation was found between carer burden and duration of illness (Zarit et al., 1980). General health status A lack of association between the GHQ-12 and carer burden, measured by the ZBI, was found in a European study (Schneider et al., 1999). Dimension-specific outcomes In the original development of the ZBI, only social support, in terms of frequency of family visits, was significantly (negatively) correlated with carer burden (Zarit et al., 1980). For the 29-item ZBI, the sense of burden was moderately to strongly associated with psychological well-being, but not significantly correlated with most variables of physical well-being (Williams 1993). One study found that high levels of burden were negatively correlated to psychological health (measured by SF-36v2) (McConaghy and Caltabiano 2005). Caregiver burden, as measured by the 22-item ZBI was not correlated with caregiver adaptability, but was correlated moderately to caregiving stress variables and memory and behaviour problems (Majerovitz 1995). Correlations for the 22-item ZBI with other measures were statistically significant, but weak, apart for depression for which a moderate correlation was found and correlations for the short ZBI were also weak to moderate (Hebert et al., 2000). Evidence of predictive validity of the short ZBI for depressive symptoms has also been found (O'Rourke and Tuokko 2003). Responsiveness A significant decrease over time in carer burden was found in wives, but not husbands, especially for wives who had placed their spouses into a nursing home (Zarit et al., 1986). ZBI scores have been shown to decrease over time, however the 373

26 decrease was greater in the waiting list group than in the intervention groups (support or counseling groups) (Zarit et al., 1987). Interpretability According to (Bedard et al., 2001), a score of 17 or above on the short ZBI (representing the top quartiles) may be used as cut off point to identify high burden. However, O Rourke and Tuokko (2003) found this not optimal, upon comparison of the short ZBI scores to scores of the Centre for Epidemiologic Studies-Depression Scale. Their suggested cut off of 10 was not optimal either, and thus it is too early to propose a definite cut off point. Acceptability Only one study reported on missing variables, with data missing only on 10 individual items (3.2%) (Hebert et al., 2000). Feasibility Both the long and the short ZBI have a low number of items. The ZBI has mostly been interviewer administered. 374

27 Table 10.16: Developmental and evaluation studies relating to the Zarit Burden Interview: Study/ Country Population (N) Age Method of administration Setting Measurement properties Zarit Burden Interview Reliability Validity Responsiveness Precision Acceptability Feasibility Zarit et al., items Caregivers or people with senile dementia (29) Age 65.0 Interviewer administered Zarit et al., item Zarit et al., items Whitlatch et al., item Williams, item Majerovitz item Husbands (31) and wives (33) who were caring for their spouse with Alzheimer s Disease Age husbands 72.3, wives 63.4 Interviewer administered Care givers of patients with dementia living in the community (119) Age 62.0 Interviewer administered Carers of non-institutionalized dementia patients (113) Age 62.0 Caregivers of stroke patients (29) Age 56.4 Spouses of patients with dementia (54) Age 70.5 Interviewer administered 375

28 Table (contd.): Developmental and evaluation studies relating to the Zarit Burden Interview: Study/ Country Population (N) Age Method of administration Setting Measurement properties Zarit Burden Interview Reliability Validity Responsiveness Precision Acceptability Feasibility Schneider et al., 1999 Europe including the UK 29-item Co-resident spouses of people with probable dementia (20) Age 71.0 Interviewer administered Hebert et al., 2000 Canada 22-item Bedard et al., , 12- and 4-item versions Canada Gallicchio et al., item O'Rourke and Tuokko, 2003 Canada 12-item version McConaghy and Caltabiano, 2005 Australia 22-item Caregivers of people with dementia in the community (327) 61.7 years Face to face interview in the caregiver s home Caregivers of cognitively impaired adults referred to a memory clinic (413) Age 61.0 Interviewer administered Carers of community-dwelling dementia patients (327) Age 61.6 Interviewer administered Carers of institutionalized and community-dwelling patients with dementia (770) Age 58.6 Interviewer administered Carers of people with dementia (42) Age 62.0 years Self-completion questionnaire 376

29 Other carer-specific instruments identified from the review The following table provides an overview of other records of carer-specific instruments identified. They have in common the fact that only one record of a study was found evaluating the instrument; insufficient evidence to justify assessing the instrument in more detail.. Nineteen single study evaluations of instruments are included. Most of the instruments were evaluated in the, only 2 in the UK. The majority of evaluations only gave information on internal consistency and validity. The Caregiver Quality of Life Instrument (CQLI) (Mohide et al., 1988) was tested more extensively, but only in a small sample. 377

30 Table General carer instruments evaluated in a single study Instrument/ reference Brief Assessment Scale for Caregivers (BASC) and Negative Personal Impact (NPI) subscale; Glajchen et al., 2005 Burden Scale; Pruchno, 1990 Population (N) Age Method of administration Setting Caregivers of patients with chronic illness (102) () Age 49.0 Interviewer administered (face to face or telephone) Carers of spouses with Alzheimer s Disease (315) Age 70.2 Intervieweradministered Reliability Validity Responsiveness Precision Acceptability Feasibility Comments Internal consistency Construct No other records identified unless stated Internal consistency was acceptable for the BASC and the NPI. Construct validity was confirmed by significant correlations to other measures of caregiver burden. (such as Objective Caregiver Burden). The burden scale was found to be internally consistent. The Burden Scale was correlated to the CES- D scale. Caregiver Activity Survey; Davis et al., 1997 Caregivers of Alzheimer patients (42) () Self-completion questionnaire Test-retest The Caregiver Activity Survey total score had high test-retest reliability. Convergent validity was supported by comparing the Caregiver Activity Survey to other Alzheimer s disease measures and an independent measure of caregiver burden. 378

31 Instrument/ reference Caregiver Appraisal Scale (CAS) (primary instrument) Subjective Burden Scale, Objective Burden Scale, General Health Questionnaire; Struchen et al., 2002 Caregiver Change Interview; Zarit et al., 1987 Craregiving Burden Scale; Knight et al., 1998 Population (N) Age Method of administration Setting Caregivers of persons with traumatic brain injury (241) Age 47.0 Self-completion questionnaire Caregivers of dementia patients living in the community (119) Age 62.0 Interviewer administered Caregivers of persons with traumatic brain injury (52) Age 47.1 Self-completion Reliability Validity Responsiveness Precision Acceptability Feasibility Comments No other records identified unless stated Factor analysis yielded 4 factors for the CAS. Three of the 4 factors showed good internal consistency. The perceived burden sub-scale had moderate correlations with the 3 other instruments, but correlations were weak for the other sub-scales (caregiver relationship satisfaction, caregiving ideology and caregiving mastery). Good internal consistency was found for the 4 sub-scales Five of the seven CBS subscales showed good internal consistency (family impact and physical burden did not have good internal consistency) For validity, Parents scored significantly higher on the pessimism and physical burden subscales of the CBS than spouses when caring for a person with traumatic brain injury. The CBS total score was significantly correlated to symptom distress, coping, social support and depression 379